Awesome

subdetection

Subtitle detection library based on Tesseract-OCR and Qt5.

Version used:

• tesseract-ocr v3.02.02
• Qt 5.1

You can check out my tutorials if you have no clue how to compile Tesseract on Windows:

http://broija.blogspot.fr/2014/10/build-tesseract-ocr-30202-with-qt-51.html

http://broija.blogspot.fr/2014/10/compile-tesseract-ocr-30202-with-mingw.html

Description

This library contains several classes, structures and functions to help detect subtitles in images.

All the parameters are tuned in the Parameters structure. You can also use a ParameterManager to save or load these parameters in a file or use the signal/slot mechanism.

The Detector class is the one to use for text detection.

Parameters

• hsvMin and hsvMax are set to define the color range in which the subtitles are printed.
• zone defines the zone were the subtitles will appear.
• charMaxSize helps the detector guess what could be a character and what is too big to be one.
• thresh is used in the contour detection process.
• xTolerance is the maximum spacing between two characters along X axis.
• yTolerance is the maximum spacing between characters on a line along Y axis.
• matchRatio represents the similariy of the "cleaned" text zones between consecutively loaded images. The closer from 0, the more similar.

Instructions

Text detection

Detector usage is quite straightforward:

• Pass it the desired parameters using one of its constructors or calling setParameters.
• Then simply call detect.

If text has been detected, you'll find it in the QStringList passed as parameter.

Language files

Tesseract language files can be used by putting them in a tessdata directory.

Then, instantiate a Detector and pass it tessdata parent directory path and the three characters refering to your language files.

You can download language files here:

https://code.google.com/p/tesseract-ocr/downloads/list

Tests

Application

A test project is available in test/app/ directory.

Unit tests

Available in test/unit_tests/ directory.